Publications

VIP: Vision tool for comparing Images of People

Michel Lantagne, Marc Parizeau and Robert Bergevin

Abstract - This paper describes the VIP technique, a Vision tool for comparing Images of People. This technique compares two human silhouettes and produces a similarity score between them. VIP was developed in the context of a surveillance project where one of the objectives is to recognize, in real-time, a person over different angles. The silhouette comparison must be robust to realworld situations, in particular to variations in scales, lighting conditions and human pose. The development of VIP involved the merging of several content-based image retrieval techniques. Colour and texture descriptors are used to describe the regions found inside a person’s silhouette and a region matching scheme associates regions of one silhouette to another. For a recall of 80%, the average precision was found to be between 76% and 93% for a 870 images database with 16 different people. These results show the robustness of the system against scale, angle, and rotation variations in human silhouette appearance.

Bibtex:

@article{Lantagne443,
author    = { Michel Lantagne and Marc Parizeau and Robert Bergevin },
title     = { VIP: Vision tool for comparing Images of People },
year      = { 2003 },
journal   = { Vision Interface },
keywords  = { Similarity Measure, Feature, Descriptor, Color, Texture, Region Matching },
location  = { Halifax },
language  = { English }
}

Last modification: 2003/04/11 by lantagne

		Home \| About \| People \| Research \| Publications \| Events \| Profile
		©2002-. Computer Vision and Systems Laboratory. All rights reserved